Projects

Sim-2-real DRL Controller Design for FCEV with Terrain Preview (Apr 2025 – Now)

This project focuses on developing a deep reinforcement learning–based controller that leverages terrain preview to optimize fuel‐cell and battery usage in fuel cell electric vehicles (FCEVs). The goal is to reduce hydrogen consumption while maintaining battery SOC within safe operating limits.

• Developed a DDPG-based control policy that uses terrain preview to proactively balance fuel-cell and battery power, minimizing hydrogen consumption while keeping battery SOC within safe limits.
• Generated H₂-vs-SOC maps across standard drive cycles and real-world conditions to guide training and validate against hardware-in-the-loop tests.
• Early results show a 10–15% reduction in hydrogen use versus no-preview strategies; ongoing work focuses on transferring the learned policy from simulation to a prototype testbed.

PI-emulating MPC controller design for Wave Energy Converters (WECs) (Jun 2022 – Apr 2025)

Efficient control is critical for maximizing energy capture from ocean waves. This project explores a feedback-tuned Model Predictive Control (MPC) framework for WECs, using LMI-based tuning for optimal performance.

• Developed a linear feedback controller and used impedance matching for gain optimization.
• Designed a constrained pseudo-PI/MPC controller, reaching < 1 % tracking error and PID-like DC gain via linear-matrix-inequality (LMI) tuning.
• Integrated an LSTM sea-state predictor for real-time gain scheduling, boosting captured power under changing wave conditions.

Linear PTO for tractor-trailer suspension system (Jan 2023 – Jun 2024)

As part of my recent work, I have focused on the development of a Linear PTO for a tractor-trailer suspension system. One of the key challenges in this project was understanding the relative movement between the Chassis and Cab of a Class 8 Commercial Tractor. To address this, I demonstrated and formulated the relative movement to better understand the system and optimize the energy extraction process. Key contributions are:

• Built a hardware prototype linear generator; measured ~80 % energy-conversion ratio across representative road profiles.
• Modeled chassis–cab relative motion and showed up to 8 kW extractable power in Class-8 operating envelopes.

AI Based Prognostics and Health Management of BLDC Motors (Sep 2018 – May 2021)

Unexpected machine failures can lead to system-wide shutdowns, reduced output, and safety risks. To mitigate this, we developed a prognostics and health management (PHM) framework for brushless DC (BLDC) motors using multi-sensor data and machine learning techniques. The system continuously monitors degradation patterns by analyzing vibration, temperature, current, and voltage signals collected under accelerated life testing. Key contributions are:

• Developed an NI-LabVIEW test rig for accelerated life testing, logging vibration, current, voltage, temperature, and speed.
• Introduced a 3rd-harmonic current feature-selection technique, improving fault-diagnosis accuracy by 10 % over conventional methods.